Image Convolver

Overview

Image convolver algorithm performs a 2D convolution operation on the input image with the provided 2D kernel. This is useful when the kernel isn't separable and its dimensions are smaller than 5x5. In other cases, it's usually preferable to use the separable image convolver algorithm due to its speed.

Input	Kernel	Output
	\[ \begin{bmatrix} 1 & 0 & -1 \\ 0 & 0 & 0 \\ -1 & 0 & 1 \end{bmatrix} \]

Implementation

Discrete 2D convolution is implemented using the following discrete function:

\[ I'[x,y] = \sum_{m=0}^{k_h} \sum_{n=0}^{k_w} K[m,n] \times I[x-(n-\lfloor k_w/2 \rfloor), y-(m-\lfloor k_h/2 \rfloor) ] \]

Where:

• \(I\) is the input image.
• \(I'\) is the result image.
• \(K\) is the convolution kernel.
• \(k_w,k_h\) are the kernel's width and height, respectively.

Note

Most computer vision libraries expect the kernel to be reversed before calling their convolution functions. Not so with VPI, we implement an actual convolution, not cross-correlation. Naturally, this is irrelevant if the kernel is symmetric.

Usage

1. Initialization phase
   1. Include the header that defines the needed functions and structures.

      #include <vpi/algo/ImageConvolver.h>

   2. Define the stream on which the algorithm will be executed, the input and output images.

          VPIStream stream = /*...*/;
          VPIImage input   = /*...*/;

   3. Create the output image.

          uint32_t w, h;
          vpiImageGetSize(input, &w, &h);
       
          VPIImageType type;
          vpiImageGetType(input, &type);
       
          VPIImage output;
          vpiImageCreate(w, h, type, 0, &output);

2. Processing phase
   1. Define the kernel to be used. In this case, a simple 3x3 edge detector.

          float kernel[3 * 3] = {1, 0, -1, 0, 0, 0, -1, 0, 1};

   2. Submit the algorithm to the stream, passing the kernel, input, output images and boundary condition.

          vpiSubmitImageConvolver(stream, input, output, kernel, 3, 3, VPI_BOUNDARY_COND_ZERO);

   3. Optionally, wait until the processing is done.

          vpiStreamSync(stream);

Consult the Image Convolution for a complete example.

Limitations and Constraints

Constraints for specific backends supersede the ones specified for all backends.

All Backends

• Input and output images must have the same dimensions and type.
• The following image types are accepted:
  • VPI_IMAGE_TYPE_U8
  • VPI_IMAGE_TYPE_S8
  • VPI_IMAGE_TYPE_U16
  • VPI_IMAGE_TYPE_S16
• Minimum convolution kernel size is 1x1, maximum is 11x11.
• The following boundary conditions are accepted.
  • VPI_BOUNDARY_COND_ZERO
  • VPI_BOUNDARY_COND_CLAMP

PVA

• Input and output dimensions must be between 65x33 and 3264x2448.
• Minimum convolution kernel size is 2x2.
• Maximum convolution kernel size is 11x11.
• Kernel weights are restricted to \(|weight| < 1\)
• Only VPI_BOUNDARY_COND_ZERO is accepted.

Performance

For further information on how performance was benchmarked, see Performance Measurement.

Jetson AGX Xavier

size	type	kernel	CPU	CUDA	PVA
1920x1080	u8	3x3	0.65 ms	0.0649 ms	0.9631 ms
1920x1080	u8	5x5	1.02 ms	0.0756 ms	1.2312 ms
1920x1080	u8	7x7	1.434 ms	0.1192 ms	1.7560 ms
1920x1080	u8	11x11	3.622 ms	0.2506 ms	3.2589 ms
1920x1080	u16	3x3	0.82 ms	0.1077 ms	1.0607 ms
1920x1080	u16	5x5	1.12 ms	0.1281 ms	1.5125 ms
1920x1080	u16	7x7	1.54 ms	0.1923 ms	2.3490 ms
1920x1080	u16	11x11	3.3 ms	0.4171 ms	4.6461 ms

Jetson TX2

size	type	kernel	CPU	CUDA	PVA
1920x1080	u8	3x3	1.5 ms	0.256 ms	n/a
1920x1080	u8	5x5	2.54 ms	0.305 ms	n/a
1920x1080	u8	7x7	4.07 ms	0.482 ms	n/a
1920x1080	u8	11x11	11.3 ms	0.7997 ms	n/a
1920x1080	u16	3x3	1.74 ms	0.381 ms	n/a
1920x1080	u16	5x5	2.83 ms	0.389 ms	n/a
1920x1080	u16	7x7	4.14 ms	0.574 ms	n/a
1920x1080	u16	11x11	11.19 ms	0.9969 ms	n/a

Jetson Nano

size	type	kernel	CPU	CUDA	PVA
1920x1080	u8	3x3	3.14 ms	0.670 ms	n/a
1920x1080	u8	5x5	5.73 ms	0.8814 ms	n/a
1920x1080	u8	7x7	8.78 ms	1.3284 ms	n/a
1920x1080	u8	11x11	25.55 ms	2.2300 ms	n/a
1920x1080	u16	3x3	3.698 ms	0.964 ms	n/a
1920x1080	u16	5x5	5.93 ms	1.0299 ms	n/a
1920x1080	u16	7x7	9.06 ms	1.564 ms	n/a
1920x1080	u16	11x11	25.92 ms	2.8124 ms	n/a